AI isn't replacing accountants - it's replacing the jobs that used to train them

AI is taking over the entry-level work that used to train accountants, not the experienced accountants themselves. Big Four entry-level openings fell 44% year over year in 2024, and for the first time the four firms are posting more AI roles than audit roles. The staffing model is moving from a pyramid to a diamond, and the skill that gets tax professionals hired now is reviewing and verifying AI output.

The Financial Times analysed 50,000 job postings from Deloitte, EY, KPMG, and PwC between 2020 and early 2026 (as reported by the Irish Times). The finding that's been circulating in accounting circles this week: for the first time, the Big Four are posting more AI roles than audit roles.

That headline is true and also slightly misleading. The more important number is this: entry-level openings across the Big Four fell 44% year-over-year in 2024. KPMG cut some graduate cohorts by nearly 30%.

AI isn't replacing the senior accountant. It's replacing the work that turned a junior into one.

What is replacing the Big Four pyramid model?

A diamond: fewer juniors at the base, a broader middle of technical and managerial experts, and AI handling the volume work underneath.

The Big Four have always run on a pyramid. Lots of junior staff doing foundational work, seniors reviewing and packaging it, partners selling it. The model worked because labour at the base was relatively cheap and relatively plentiful.

That base is being automated. KPMG already uses AI to scan millions of accounting entries and flag exceptions for human review. EY's AI assists 80,000 tax professionals and handles over 3 million compliance cases a year. The four firms spent over $9 billion building this infrastructure.

The result is what Olivier Khatib, who wrote a widely read analysis of this shift, calls a "diamond model": a thinner base, a broader middle of technical and managerial experts, and AI handling the volume work underneath.

What do the new Big Four AI roles require?

Most require programming skills, but a meaningful share are non-technical roles for people who can check that AI output holds up.

The FT only counted roles where working with machine learning or generative AI was a core requirement, not vague "digital transformation" titles. Specific examples included a KPMG role for a manager directing AI to automate tasks via prompt engineering, an EY senior associate helping clients integrate AI into their tax functions, and a Deloitte position requiring eight years of AI strategy experience.

Around 80% of these postings required programming expertise. But a meaningful share were non-technical: product managers overseeing AI audit tools, and people who could sit between the model and the client and verify that outputs held up. That last category is the one most experienced tax professionals are best positioned to fill.

What does this mean for tax professionals?

The entry point to the job moves from doing the data entry to reviewing what AI produced.

Two things are both true right now. The accounting industry will likely grow, because cheaper and faster services create more demand. And the skills that get you hired are changing fast.

The junior who used to learn by doing the data entry now needs to learn by reviewing what AI produced. That's a different skill, and most firms aren't training for it yet. The ones that do will have a structural advantage that compounds every tax season.

For anyone earlier in their career: the credential that matters in 18 months is not whether you can use AI tools. It's whether you can audit their output, catch failure modes, and explain your review process to a partner or a regulator.

We built the AI Tax Specialist and Engineer course for exactly that. Free, self-paced, seven modules, CPE eligible.

Enroll for free in the AI Tax Specialist and Engineer course

Frequently asked questions

Is AI replacing accountants?

No. AI is taking over high-volume, repetitive work such as data entry and first-pass testing, while judgment, review and client work stay with people. The change shows up in hiring: fewer entry-level openings and more demand for professionals who can check AI output.

What is the diamond model in accounting?

A staffing structure with a thinner base of junior staff, a broader middle of technical and managerial experts, and AI handling volume work underneath. It replaces the traditional pyramid, which depended on large graduate intakes doing foundational work.

How does Filed fit into this shift?

Filed does the analytical prep and review work so tax professionals can spend their time on judgment. Filed Prep reads the source documents and prepares the return across 1040, 1065 (with K-1), 1120, 1120-S, 1041, and 990 returns, inside CCH Axcess, UltraTax, Lacerte, Drake, and ProConnect. Filed Reviewer then checks the draft against those documents line by line before a named reviewer at the firm signs off.

What skills should junior tax professionals build now?

Reviewing AI output: auditing it against source documents, catching failure modes, and explaining your review to a partner or regulator. Our guide to AI tools for tax professionals covers where these tools fit today.

Sources

• Financial Times - original analysis (may require subscription)
• Irish Times - Big Four post more AI job ads than auditor ads
• Computerworld - Big Four now hiring more AI specialists than accountants
• Scottish Financial News - Big Four slash graduate jobs as AI takes over
• TheStreet - Big Four chooses AI over humans, cuts benefits and hiring
• Business Insider - 4 key ways AI changed the Big Four in 2025
• Olivier Khatib - AI and the collapse of the Big Four?